Heterogeneity of beta-cell function in subjects with multiple islet autoantibodies in the TEDDY family prevention study - TEFA

Abstract

Background: Individuals with multiple islet autoantibodies are at increased risk for clinical type 1 diabetes and may proceed gradually from stage to stage complicating the recruitment to secondary prevention studies. We evaluated multiple islet autoantibody positive subjects before randomisation for a clinical trial 1 month apart for beta-cell function, glucose metabolism and continuous glucose monitoring (CGM). We hypothesized that the number and type of islet autoantibodies in combination with different measures of glucose metabolism including fasting glucose, HbA1c, oral glucose tolerance test (OGTT), intra venous glucose tolerance test (IvGTT) and CGM allows for more precise staging of autoimmune type 1 diabetes than the number of islet autoantibodies alone.

Methods: Subjects (n = 57) at 2-50 years of age, positive for two or more islet autoantibodies were assessed by fasting plasma insulin, glucose, HbA1c as well as First Phase Insulin Response (FPIR) in IvGTT, followed 1 month later by OGTT, and 1 week of CGM (n = 24).

Results: Autoantibodies against GAD65 (GADA; n = 52), ZnT8 (ZnT8A; n = 40), IA-2 (IA-2A; n = 38) and insulin (IAA; n = 28) were present in 9 different combinations of 2-4 autoantibodies. Fasting glucose and HbA1c did not differ between the two visits. The estimate of the linear relationship between log2-transformed FPIR as the outcome and log2-transformed area under the OGTT glucose curve (AUC) as the predictor, adjusting for age and sex was - 1.88 (- 2.71, - 1.05) p = 3.49 × 10-5. The direction of the estimates for all glucose metabolism measures was positive except for FPIR, which was negative. FPIR was associated with higher blood glucose. Both the median and the spread of the CGM glucose data were significantly associated with higher glucose values based on OGTT, higher HbA1c, and lower FPIR. There was no association between glucose metabolism, autoantibody number and type except that there was an indication that the presence of at least one of ZnT8(Q/R/W) A was associated with a lower log2-transformed FPIR (- 0.80 (- 1.58, - 0.02), p = 0.046).

Conclusions: The sole use of two or more islet autoantibodies as inclusion criterion for Stage 1 diabetes in prevention trials is unsatisfactory. Staging type 1 diabetes needs to take the heterogeneity in beta-cell function and glucose metabolism into account.

Trial registration: ClinicalTrials.gov identifier: NCT02605148 , November 16, 2015.

Keywords: Continuous glucose monitoring; Glucose metabolism; Islet autoantibodies; beta-cell function.

Conflict of interest statement

The authors declare that they have no competing interests.

© 2022. The Author(s).

Figures

Fig. 1

The distribution of fasting glucose (mmol/L) (1A) and HbA1c (mmol/mol) (1B) at visits 1 and 2 (n = 57). The boxplot indicates the median, the interquartile range, the violin around the boxplot shows the shape of the distribution of values at each visit, with individual data points shown in grey. The fasting glucose was measured 10 min before the start of IvGTT at visit 1 and OGTT at visit 2. The dashed lines at 7 mmol/L in 1A and at 42 mmol/mol in 1B indicate the World Health Organization thresholds above which subject is considered to be diabetic. Fasting glucose measurements were not found to differ between visits 1 and 2, nor did HbA1c at visits 1 and 2 (Wilcoxon test p-value 0.99 and 0.27, respectively)

Fig. 2

Assessment of glucose metabolism based on OGTT (2A and B), IvGTT (2B) (n = 57) and CGM (2C) (n = 24). In panel 2A we show the individual trajectories of blood glucose measured using a 2-h OGTT test at visit 2. Based on glucose values at minute 120, we identified 7 subjects (labelled with red and orange subject labels) would be considered to have impaired glucose tolerance (orange labels) or to have clinical type 1 diabetes (red labels) according to the World Health Organization (see grey panel in 2A). In panel 2B we show a scatterplot and a regression line between the log2-transformed area under the curve (AUC) of OGTT glucose measurements versus the log2-transformed FPIR measurements from IvGTT at visit 1. The subjects labelled in orange and red correspond to those in panel 2A. In panel 2C we present the distributions of the glucose measurements obtained from the Continuous Glucose Monitor (CGM) over a 7-day period starting at visit 2. The subjects were sorted according to an increasing median glucose value. The individual glucose measurements are shown as points, with the boxplots showing the median and interquartile range, and the violin plot showing the distribution of the CGM glucose values for a given individual. The subjects shown in orange and red correspond to those in panels 2A and 2B

Fig. 3

FPIR (mU/L) levels and autoantibody status. The distribution of log2-transformed first-phase insulin response (FPIR) (mU/L) stratified by presence of IAA, GADA, IA-2A or ZnT8(Q/R/Q)A (n = 57). The boxplot indicates the median, the interquartile range, the violin around the boxplot shows the shape of the distribution, with individual data points shown in grey. Based on the model in Table 3 with the status of the four antibodies as the main predictor, adjusted for age and sex, log2(FPIR) was not found to be statistically significantly different depending on the status of IAA, GADA or IA-2A (p = 0.564, 0.637, 0.262), but there was some evidence suggesting that FPIR is lower for those with at least one of ZnT8(Q/R/W)A present compared to those with no ZnT8(Q/R/W)A (1.74 mU/L lower (95% CI = 1.01, 2.99), p = 0.046)